Don't be shellfish...Share on FacebookTweet about this on TwitterShare on Google+

We’ve posted the first version of 2010 Razzball Point Shares (also available via the 2010 Fantasy Baseball Rankings header).  For end of year 2009 Point Shares, please click here.

This first version is based solely on CHONE projections except for Saves which we added (CHONE doesn’t estimate saves).

In late February/early March, we’ll add ZiPS projections and reweight the projections based on the latest playing time expectations.  If a certain player looks high/low, you should consider the plate appearances.  For example, you’d be hard-pressed to find anyone ranking Orlando Cabrera ahead of Jason Bartlett.  But Cabrera is projected at 87 more plate appearances (631 vs. 544) which helps pad his R/HR/RBI/SB totals.  The impact that playing time can have is even more dramatic when looking at starting pitcher value.

For those unfamiliar, Point Shares are our home-grown methodology (inspired by Bill James’ Win Shares) for rating fantasy players.  Besides providing a basis for rating players, Point Shares provide quick estimates as to a player’s impact to your overall team points and compare across categories – e.g., Ichiro’s contributes 1.3 points in SB vs. the average OF and -1.3 points in HRs vs. the average OF.

Here are the basics:

  • Point shares represent a player’s estimated impact on an average team’s points versus the average drafted player at his position.  So in a 12 team league, the ‘average drafted SS’ would be between the 9th and 10th valued SS (12 starting SS, 6 slotted for 2B/SS) and the average team would get 65 points (6.5 per category if 12 is top and 1 is bottom).    For example, Hanley Ramirez’s 9.69 Point Shares mean he was worth 9.69 points more to the average team than the average stats for SS in a 12-team MLB league and move the team up to 74.69 points.
  • Point Shares are estimated for each statistical category.  Hanley Ramirez’s 9.69 Point Shares break out as: 2.4 Runs, 1.7 HRs, 1.0 RBI, 2.5 SB, and 2.1 AVG.
  • The assumed roster makeup is C/1B/2B/SS/3B/CI/MI/5 OF/Util/9P.  The assumed number of SPs and RPs per team are 5.5 to 3.5.  The UTIL slot is manned by DH and some OF, 1B, and 3B.
  • Players are ranked based on the position in which they are judged most valuable and are 20 games eligible. The order of most valuable to least valuable for positions is: C, SS, 2B, 3B, OF, 1B, DH.
  • Player value is primarily driven by positional comparisons (75% for hitters, 85% for pitchers) with the remained based on the average hitter/pitcher.  Batting Average, ERA, and WHIP are adjusted based on a player’s ABs/IPs (e.g., the more ABs/IPs, the greater a positive/negative impact it has for a fantasy team).

Here are some answers to some common questions:

1) Why do SPs rank higher in Point Shares than they are typically drafted? (e.g., 10 of top 20 in final 2009 PS, 2 of top 3 for 2010)

The best pitchers generally dominate 4 categories (W, K, ERA, WHIP).  While they don’t contribute in Saves, there is little opportunity cost to this since saves are concentrated in 30-40 players and there are 90+ pitching roster spots (108 for 12 team).  For hitters, it’s rare for a player to be excellent across more than 3 categories and there’s a larger opportunity cost for the categories they are not great (i.e., OFs like Ichiro and Taveras hurt HR/RBI).  Note that only 8 players in 2010 are projected to be average or better for their position in all 5 stats (Hanley, Pujols, Braun, Wright, Holliday, A-Rod, Kemp, Beltran).

Here’s another way of looking at it.  If you converted Zack Greinke’s 2009 season to an OF (W=R, HR=SO, RBI=WHIP, SB=SV, AVG=ERA), his line would have been 103 /40/108/11 /.367.   Lincecum and Felix Hernandez weren’t too far off this.  The reason why it’s usually not advisable to draft a pitcher in the 1st round is the unpredictability of who’ll be the top pitchers vs. their true value.  Here’s a more thorough explanation of a pitcher’s fantasy value.

2) Why are Point Shares based off value over an average player versus a replacement player?

There are several ‘real’ baseball metrics that value a player against ‘replacement players’ including Baseball Prospectus’ VORP and Sean Smith’s WAR.  These take position into account as the offensive value of a replacement-level 2B is going to be less than than an OF.

Point Shares uses ‘average player’ for two reasons:

a) ‘Replacement Player’ in fantasy baseball doesn’t properly estimate positional value

Fantasy baseball has two key differences that converge the replacement level for all positions to a similar offensive value:  1) The number of players selected per position varies in fantasy baseball (e.g., 1 C / 5 OF) vs. real baseball (1 C / 3 OF) and 2) The Utility position which disproportionately grabs 1B/OF given their greater hitter depth.

Pos MLB 12-Team Utility % Used
C 30 12 0 40%
1B 30 18 2 67%
2B 30 18 0 60%
SS 30 18 0 60%
3B 30 18 1 63%
OF 90 60 6 73%
DH 5 0 3 60%

Note:  UTIL distribution is estimated

The below chart shows how the value of hitters converges across all positions by the time you reach  ‘Replacement Value’ or, in other words, the Free Agency wire.  If there are any inequities where a certain position has greater depth, the UTIL slot will exploit it (and that doesn’t even account for Bench).  So if one used a true ‘replacement value’ estimate by position, there would be little to no positional adjustment despite the fact that, all statistical equal, a player with C, 2B, or SS eligibility is more valuable than someone with 1B or OF eligibility.   By using ‘average’ player value, the differences in offensive value among the top-end players can be taken into account providing a better reflection of a player’s value vs. the players owned on your competitors’ teams.

Note:  Based on 2009 data with value estimated by using Tom Tango’s formula of HR + SB + (H – .27*AB) + R/3 + RBI/3

b) Point Shares estimates a player’s impact on the average team vs. their impact on a replacement-level team.

Let’s say it’s draft day in a 12 team MLB league and your team looks like a keg-league softball team.  All power, no SBs.  There is no doubt that, should you keep that roster, you will finish first in HRs and last in SBs.  If you trade Ryan Howard for Jacoby Ellsbury, you may see no impact in HR/RBI and gain 6 points in SB.  If the other team was built like a Herzog-era Cardinals team, they might see a lift in HR/RBI with little to no impact on SBs.

These are extreme cases of course.  It’s rare that a random team is destined to be in last place for a statistic (random meaning that we’re not talking about a specific team that was on auto-pilot for draft day or is managed by that doofus in your office) .  So let’s say the odds are 11-1 a team in a 12-team league would finish last in a statistic.  It wouldn’t make sense to value Ellsbury’s SBs the way that the most SB-desperate team would value them, right?

Well, that would be precisely the way a ‘replacement-level’ methodology would look at it.

If we look at things from an average perspective, you would give equal probability that a team would finish in 1st place through 12th place for each stat.  If the best team is at 120 points and the worst at 10 points, the average team is at 65 points ((120+10)/2)  or 6.5 points per category.  For a team to have 6.5 points per category, we need to assume they are average at every position vs. has a replacement-level player.

3) What is the highest/lowest score possible?

Theoretically, the best/worst possible Point Shares score for a category would be +/- 5.5 pts in each category (moving someone from 6.5 to 12 or 1).  Any other league size, just divide the team total by 2 and subtract 0.5 (10 team = 4.5).  A simpler way of looking at it would be to consider how one would calculate the average value per team of a player with 50 SBs.  For the team that’s 1st in SBs, his value is 0.  For the 2nd team, it could be worth +1.  Take this to its logical conclusion and you get (0+1+2+3+4+5+6+7+8+9+10+11)/12.  That’s 66/12 = 5.5.

Note:  The highest points within a single category in 2010 Point Shares v1 is Jacoby Ellsbury’s 3.6 points in SB.

——-

Follow me on twitter:  rudygamble
Follow me on FB:  Rudy Gamble

  1. Orion says:
    (link)

    Awesome. Are there any plans to put these values into your Draft War Room sheets?

  2. Will have to check w/ our War Room creator Lou but I don’t see why not.

    I’d hold off on adding this version as adjusting playing time leads to significant changes. I think 1st/2nd week of March should be just about final…

  3. Orion says:
    (link)

    Good point, no use integrating them until they’re more accurate. I’ll hound Lou then :)

  4. Joel says:
    (link)

    Rudy,

    A couple of questions.

    Where are you deriving your point values from? Obviously league layout has a big impact on how much points are worth.

    Another; do you think that the FANS projections from fangraphs will be useful? I like them because I think the fans will nail playing time more accurately without my having to tweak anything (unlike the completely automated projections systems).

    Yes, they’re optimistic, but maybe they can be leveraged against CHONE or ZiPS.

  5. @Joel: It gets a little complicated. Here it goes. The points are based on a 12-team MLB league. I take the highest value players to create virtual rosters and total the stats by position to create position averages. I take the total data per category and assign a % distribution based on this total to determine the increment that equals a point. For instance, HR for 12-team ends up being estimated at 8 HRs by point based on the CHONE stats. So a player has to hit around 8 HRs above position average to be credited with 1 Point Share in HRs.

    I am keeping an eye on the FANS projections. I think they’ll end up more accurate on a total basis (e.g., player x hits 30 HRs vs 26) but that’ll mostly be driven by more educated playing time assumptions. I think CHONE and ZiPS are still better at determining the correct rates (e.g., HR/Plate Appearance) and end up applying those rates to playing time data (which I prefer using Baseball Prospectus because it ensures that the total Plate Appearances and IP total up to the expected total for 30 teams/162 games).

  6. GTS says:
    (link)

    How do I adjust for a 10-team league?

  7. Pins on Shins says:
    (link)

    Joe Mauer…#7. Interesting.

  8. @GTS: The point shares average about 15% less as there are 1/6 less teams (2 divide by 12). The rankings would generally stay the same but there would be larger variations based on position depth (e.g, the deeper a category, the less valuable a player is).

    Will consider doing a 10 team version for version 2…

  9. @Pins on Shins: I know….I’d never take him in the 1st round but catching is definitely the position with the weakest depth. He did finish 6th last year despite missing the first month…

  10. Rabbit says:
    (link)

    Rudy, are the player auction values listed on your Point Share sheets derived from your Point Share valuations, or do they come from somewhere else? If there derived from Point Shares, how do you get them?

  11. brad says:
    (link)

    a couple of my observations, these probably have much more to do with CHONE’s projections than the Point Share evaluation:

    CHONE must like Vazquez in NY more than I do. #23 and $28 seems pretty high.

    BJ Upton about 30 spots and $5 higher than Justin.

    Evan Longoria way down at 64.

    Aaron Hill 110 just above Francoeur. Ouch.

  12. @Rabbit: Yes, they are derived from the Point Share valuations via a formula – PS / 0.25 + 12. I developed it a year or two ago – when I tested it, the $ sum for a fully rostered league came very close to $260 * 12.

    @brad: Yes on Vazquez – CHONE projects him at modest ERA/WHIP (3.83/1.22) but 14 Wins and 182 Ks in CHONE ranks 3rd (Halladay – 16, CC – 15) and 6th respectively.

    Interesting catch on the Uptons. BJ gets his greatest value in 40 projected SBs. Justin Upton’s stats are just average to slightly above average across the board: 73/22/72/13/.296. Note that this is with 511 PAs / 456 ABs which is on the low side in my opinion.

    Longoria – yes, surprising, especially b/c he finished #24 on PS last year.

    Hill – His stats look pretty good in CHONE but I think his PS are hurt b/c he brings no speed (-0.9 PS) to a position where 6 of the top 7 are projected at 12+ SBs (and SBs are about a point in the standings for every 10 SBs).

    Francouer – Disregard. He’s projected at 631 PAs which spikes up his R/HR/RBI. He’s lucky to get 500 PAs (which, for him, means 495 ABs).

  13. matthole says:
    (link)

    Wright ahead of both Ar0d and Longoria (#65?) ….pass on longoria rd 1/2?

  14. @matthole: I’m not buying Wright at that slot at all. I cross-reference ZiPs projections for Wright and they are very close to Zips – 90+ Runs and RBIs, 20+ HRs and SBs, .300+ AVG. I’d say A-Rod is the only one of the three I’d seriously consider for a top 5 draft pick. I wouldn’t pick Longoria in 1st round but definitely a case for 2nd round (he finished #24 in PS last year)

  15. brad says:
    (link)

    @Rudy Gamble: I’m ok with the wins and Ks projections, but the Vazquez ERA and WHIP are probably low.

    I took a look at J-Up’s CHONE projections after I commented and , a slight regression in every stat from last year. Which is certainly possible, but it’s more likely we see a progression, and I think it’s worth paying for the upside. My philosophy last year was that I’d rather be a year early on J-Up than a year late – as good as he was last year, I think there’s room for improvement and I guess I’d rather overpay for 73/22/72/13/.296 than miss out on that. As for BJ, 40-50 seems okay to me, I just thought it was interesting that he was ranked ahead of his bro.

    Longo’s PS appear to be hurt by the projected PAs as well. I see no reason to think he won’t see around as many as Pujols, yet Albert’s projected to get about 100 more (as is Jeff Francouer strangely.) I would imagine the CHONE projections will normalize as we get closer to April.

  16. @brad: I think Vazquez can put up that ERA/WHIP. I think he’s a solid bet for a top 12 starter this year.

    I do agree that Upton at those projections is more valuable than a 32 OF with those projections given the upside. Sometimes gambling on upside is what you need to win a league. I tend to make my upside bets in the 2nd half of drafts vs. the 1st half though.

    CHONE just isn’t build for projected PAs. That’s why I’ll just apply the PA rates to a different PA source like Baseball Prospectus.

  17. Milkman of Doom says:
    (link)

    If we can merge these stats with the War Room sometime in mid-march that would be incredible. Please make it happen I promise to click on all this site’s ads every day for a year.

  18. Rabbit says:
    (link)

    @Rudy Gamble: I’m surprised you’re bullish on Vazquez this year. Why wou;dn’t you think he would regress to numbers more like his previous stint with the Yanks, or at least his years with the White Sox? Moving to a hitter’s ballpark in the AL just doesn’t seem like a good deal for this guy.

  19. Howie says:
    (link)

    @Rudy Gamble: Would like to petition for v. 2 to be geared around an 8-team league. Thanks.

  20. @Rabbit: I don’t think Vazquez will be as valuable as last year but a 3.80/1.29 with 15 W and 200 Ks makes for a solid 2nd starter.

    @Howie: 8 team league? MLB? You need more competition. On that note, look out for the Razzball Commenter League signups. coming soon…

  21. Nick J says:
    (link)

    Didn’t I have a bet with some guy about Vazquez on a thread here last year?

  22. Cygar says:
    (link)

    I love this spreadsheet. Very informational and thoroughly researched.

    Question for you about a person in the early 100′s.

    Eric Young Jr.

    What in the 10 games he played last year (and not starting spot availabe this year) did you see to have him ranked so high??

    Ahead of Ben Zobrist, Gordon Beckham, Derrek Lee, Aramis Ramirez, Pablo Sandoval, etc etc etc.

    Should there be some guys in there that we just acknowledge that they are misplaced??

  23. @Nick J: Hmm, not sure but I know Vazquez got a lot of chatter on Razzball last year after the Point Shares based on CHONE had Vazquez in the top 5 overall. Sean – the guy behind CHONE – was even surprised by how well Vazquez’s projects stats looked. But you know where Vazquez finished in Point Shares – #7! (http://razzball.com/2009-fantasy-baseball-rankings-final/2009-fantasy-baseball-player-rater-end-of-year-point-shares/). Based on CHONE’s 2010 projections for JV, it’s clear that there’s a healthy adjustment going from NL East to AL East. But he K’s a lot of guys, has been healthy for a number of years, and doesn’t have a huge BB rate. There’s a lot of value there. Not enough that I’d consider him an ace but he’s a solid #2.

    @Cygar: Just so you know, I don’t fudge with any player rankings based on personal opinion. I wouldn’t use Point Shares – even the final version – as a sacred draft sheet. I was incredibly shocked to see Eric Young Jr. valued ahead of those players. I was really surprised he’s ahead of Aaron Hill at 2B. As you can see by the Point Share breakout by category, his value is driven mainly by Runs and SBs. CHONE has him at 99 runs and 38 SBs. That’s 7th overall in Runs and 5th in SBs. With 2B eligibility, switch-hitting, and near .300 average, he’d be a reincarnation of Luis Castillo at his peak. I checked ZiPS (http://www.baseballthinkfactory.org/files/oracle/discussion/2010_zips_projections_colorado_rockies/) and it projects 84 runs and 45 SBs BUT a .257 average.

    I think the biggest question mark on Eric Young right now is his place in the order. He’s shown solid OBP skills in the minors (.387 and .391 last two years) that, combined with his speed, makes him a potential #1/#2 hitter this year (albeit with OBP regressing to the .340/.350 range).

    These are the guys who hit #1/#2 for the Rockies last year:
    Fowler (102 games)
    Barmes (62 games)
    Sean Smith (49 games)
    Spilbourghs (35 games)
    CarGo (40 games)
    TuLo (~ 20 games)

    Assuming they use TuLo and CarGo somewhere b/w 3rd and 5th in the order, I can definitely see Eric Young in the #1/#2 slot. My concern at this point is that Barmes is still on the roster and, at the very least, will play 40 games or so at 2B. And if Young gets only 400 ABs, he’ll still get 30 SBs but his Runs won’t look so impressive.

    But I’ve definitely got EYJr. on my sleeper list now…

  24. Cygar says:
    (link)

    Being a Colorado resident, I love me some EYJ.

    I just hope he can get the AB’s. As of now he is destined for Colorado Springs. With Barmes and Mora manning 2B and Fowler, CarGo, Smith, Spilly and Hawpe in the OF, I just see no spot for him.

    CarGo went nuts last year in Colorado Springs and still only got 278 AB’s with the parent club. Seems like the Rockies aren’t playing them as young as they used to because they are a quality club now.

    While I think EYJ would be a great table setter for the big bats in the Rockies line up and an obvious lead off hitter for them, I just don’t see them benching Barmes and Mora because of experience and money.

    How many AB’s do you see for EYJ? I am guessing around 200. That is why I am leaning more towards Everth Cabrera, Brantley, Stubbs, and Juan Pierre as later round speedsters because they have spots right now.

    I put EYJ in the Desmond Jennings category, just waiting for a spot to open.

  25. Joel says:
    (link)

    Rudy,

    Just some more thoughts.

    1) I’ve been leaning towards using a baseline automated projection (if you use BP’s playing times with ZiPS and/or CHONE, that’ll probably be it!) but with FAN projections as my “upside”. Since they’re almost unilaterally optimistic, they can give me an idea of who might really break out, without having to pore through PECOTA’s breakout percentages (or having to pay for them).

    2) A crucial aspect of drafting is knowing when your compadres will pick players. In the past, I took MDC’s ADPs as my “mean draft positions” and then estimated standard deviations by taking the range (lowest – highest pick) and dividing by 4 (this is a pretty standard estimating practice and uncannily accurate). I could then figure out the probability that a player would be available at a given slot by using Excel’s NORMDIST function.

    For people who still sit in a room, drinking beer, and drafting, this is pretty handy. It let me sit and wait on some players I really wanted (but didn’t want to reach for) like Vasquez and Bruce. On the other hand, I whiffed on Reynolds (got him later on the waiver wire) and a few others.

    The problem with this system is that

    a) The MDC data is not very good. But they’re the only ones that I know of that provide ADP and ranges.

    b) Distributions of picks aren’t actually normal. For one, there’s a huge positive skew for high picks (i.e. almost everyone is going to pick Pujols first and Ramirez second). There’s also a huge negative skew for low picks (some players that might get picked as high as the 14th round (long tail) are far more likely not to get selected at all). Even more complicated are bimodal distributions. A draft with 12 dorks like myself is going to entail a lot of people making similar picks. In a heterogenous league, you’re going to see a lot of variation. We had a lot of “sleepers” going early (like Chris Davis) but Ichiro slid to the 6th round and Mauer slid to the 9th. On the other hand, Mark Reynolds lasted until the 14th, and Greinke until the 13th.

    Anyhow, I’m wondering what your thoughts are on this method. In fact, I think with an automated program like the war room you could really get a lot of statistical strength here by adjusting percentages based on who’s already been taken. You’d need access to a lot of raw mock draft data to make it work, though.

  26. The Dude says:
    (link)

    @Rudy Gamble: Since all my leagues are “Yahoo Style’ (i.e. no CI, MI and only 3 OF), how does this affect theses point shares? Can it be adjusted for a Yahoo league?

    Thanks

  27. @Cygar: Thanks for the intel. EYJ seems like a good sleeper for NL-only but is unusable until he gets a starting spot.

    @Joel: I like the way you think. I tend to use MDC ADP but it’s a better representation of ‘advanced’ players’ picks vs. average players. For average players, ESPN or Yahoo rankings work best.

    I take the ADP and determine the round it would fall in. One round early is a slight reach. One round late is a slight bargain. More importantly, though, I’m studying everyone’s rosters. If I need a 2B and several teams do too, I’m more likely to reach for a 2B I like. If most teams seem set on 2B and MI, I’ll hold off.

    In other words, my strategy is to pick all players between my value and when a player would theoretically be taken in the draft (which will vary from ADP based on the flow of each draft).

  28. Howie says:
    (link)

    @Rudy Gamble: NL-only for the 8-team Roto league I am in…

  29. @Howie: Ah, that makes more sense. I’m going to do an NL-only for Round 2 when there are realistic playing time estimates. Even more important for NL-only since it’s a deeper draft. I was thinking 12-team vs. 8-team though. Let me think about it – maybe i can program the calculations to just knock out a bunch of league combinations….

  30. Paul says:
    (link)

    Don’t for a moment believe in the value of “win shares”, but I will say that in the one 18 team roto league, I had Mauer, Reynolds, and Greinke, and then picked up Verlander after his owner dropped him. Worked out well.

  31. Giacomo says:
    (link)

    Is there anyway to make these rankings customizable? I am specifically interested in changing the positions considered (more of a Yahoo style league) and replacing AVG with OBP.

  32. @Giacomo: Sorry. I don’t have the calculations set up to handle OBP. I played in one OBP league last year and just adjusted the AVG Point Shares up/down based on their BB rate.

  33. Fenris-77 says:
    (link)

    Based on the current PS info you have up shouldn’t Utley be added to that group of guys who are average to above average to better for their position in all five categories? Maybe I’m reading it wrong…

  34. Dave says:
    (link)

    Rudy,

    I’ve been working on my own version of this without knowing what you were up to until just now. This is ambitious and cool.

    Here is what I really like about your approach:

    (1) You build your “points universe” out of the projected statistics instead of expected statistics for 2010. That is, instead of looking at last year and saying “it took 1100 runs to finish third in the category,” you determine this from the projections themselves. I’m finding this to be really important — essential, even.

    (2) You assign players to a “most valuable” position if there are multiple positions they count toward. I agree completely.

    Still trying to wrap my head around using average performance instead of replacement performance as a baseline. I’ll reread that a few more times and see if I can make heads or tails of it.

    One thing I built into mine, and something that I would love to see in yours, would be a way to credit players with low playing time projections. For instance, Chipper Jones. I don’t leave my 3B spot empty when Chipper Jones makes his annual trip to the DL. I fill it with a replacement player. So the “net” value of drafting Chipper Jones is really 120 games of Chipper (his full projection) plus 40 games of replacement level 3B.

    Awesome work. Is it available for download as a spreadsheet?

  35. Eric Kuflik says:
    (link)

    I am sorry if this is a re-post, my other message did not make it through. If you ranked against replacement, then a pitchers value would go down, as it should. This is because we play more batter than pitchers. A replacement level batter (the worst batter starting on a team), will have a lower point share total that the replacement SP. When you add this value to the player, as you should, then SP would no longer rank as high as they do.

  36. Mike says:
    (link)

    Rudy, why do you only base 75%/85% of the point share value on positional comparisons? The rest of your methodology looks very good, but that piece seems arbitrary.

Comments are closed.